ECE: Electrical & Computer Engineering
ECE News

Cognitive radios that perform well while learning

Wireless @ Virginia Tech

Read more about ECE’s work in the wireless field at the Wireless @ Virginia Tech website.

A cognitive radio that determines its own best communications method for a situation is closer to reality, thanks to recent results from an ECE wireless communications research team.

Postdoctoral associate Haris Volos (Ph.D. ’10) and associate professor R. Michael Buehrer have developed cognitive radio techniques that enable a radio to perform well while learning and selecting its optimal configurations.

Typically, radio designers analyze each communication method in terms of goals and will arrive at a set of adaptation rules for the radio, according to Volos. “The analysis is time-consuming and if the channel models used do not hold, or in an unexpected situation, the design becomes irrelevant,” he says. Cognitive radios using a cognitive engine (CE) have been proposed that find the optimal configuration. However, learning speed and performance during learning have been big challenges, he says.

Funded by an NSF grant, Volos and Buehrer have demonstrated if there is a moderate number of methods that meet minimum performance requirements, that a CE can reach near-maximum performance in a relatively short number of trials. They have also developed a Robust Training Algorithm (RoTA) that attempts to maintain a minimum performance level while trying possibly better-performing methods.

“Our training algorithm makes the link stable during learning,” Volos says.

“A soldier attempting to use a radio during a critical operation would rather have a slowly improving connection that a connection that erratically fluctuates,” he explains. “Likewise, a mobile user streaming a favorite show prefers a slowly improving connection than a link that pauses at crucial moments.”

The work was honored in November with the Fred Ellersick Award for Best Paper in the Unclassified Technical Program at MILCOM 2010.